RESUMO
Aiming at the shortcomings of the traditional sparrow search algorithm (SSA) in path planning, such as its high time-consumption, long path length, it being easy to collide with static obstacles and its inability to avoid dynamic obstacles, this paper proposes a new improved SSA based on multi-strategies. Firstly, Cauchy reverse learning was used to initialize the sparrow population to avoid a premature convergence of the algorithm. Secondly, the sine-cosine algorithm was used to update the producers' position of the sparrow population and balance the global search and local exploration capabilities of the algorithm. Then, a Lévy flight strategy was used to update the scroungers' position to avoid the algorithm falling into the local optimum. Finally, the improved SSA and dynamic window approach (DWA) were combined to enhance the local obstacle avoidance ability of the algorithm. The proposed novel algorithm is named ISSA-DWA. Compared with the traditional SSA, the path length, path turning times and execution time planned by the ISSA-DWA are reduced by 13.42%, 63.02% and 51.35%, respectively, and the path smoothness is improved by 62.29%. The experimental results show that the ISSA-DWA proposed in this paper can not only solve the shortcomings of the SSA but can also plan a highly smooth path safely and efficiently in the complex dynamic obstacle environment.
RESUMO
To address the problem that large pedestrian detection networks cannot be directly applied to small device scenarios due to the heavyweight and slow detection speed, this paper proposes a pedestrian detection and recognition model MobileNet-YoLo based on the YoLov4-tiny target detection framework. To address the problem of low accuracy of YoLov4-tiny, MobileNetv3 is used to optimize its backbone feature extraction network, and the MFF model is proposed to fuse the output of the first two layers to solve the information loss problem, and the attention mechanism CBAM is introduced after strengthening the feature extraction network to further improve the detection efficiency; then the 3 × 3 convolution is replaced by the depth separable convolution, which greatly reduces the number of parameters and thus improves the detection rate, then propose Ordinary data augmentation to efficiently augment the dataset and dynamically adjust the target detection anchor frame using the k-means++ clustering algorithm. Finally, the model weights trained by the VOC2007 + 2012 dataset were applied to the pedestrian dataset for retraining by the transfer learning method, which effectively solved the problem of scarce samples and greatly shortened the training time. The experimental results on the VOC2007 + 2012 dataset show that the average means accuracy of the MobileNet-YoLo model compared to YoLov4-tiny, MobileNet-YoLov4, MobileNet-YoLov3, and YoLov5s by 5.00%, 1.30%, 3.23%, and 0.74%, respectively and have reached the level to realize the landed application.